Intra Sub-partitions Video Coding Parallel Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems face challenges in efficiently predicting sub-partitions for intra coding, particularly in scenarios where the prediction of one sub-partition is independent of the reconstruction of a neighboring sub-partition.
Innovation Solution
The implementation of intra sub-partitions (ISP) allows for the prediction of sub-partitions using prediction samples from neighboring sub-partitions, and enables multiple sub-partition based prediction where intra-predictions of two or more sub-partitions are conducted simultaneously.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional intra coding predicts sub-partitions sequentially using reconstructed samples from previous neighboring sub-partitions, then prediction accuracy is maintained, but processing time and latency increase
Solution Approach 1:
The patent divides the prediction process into independent sub-partition units that can be processed simultaneously. Each sub-partition is predicted using prediction samples from neighboring sub-partitions rather than sequentially using reconstructed samples, enabling parallel processing while maintaining prediction accuracy.
Solution Approach 2:
The patent performs preliminary prediction of multiple sub-partitions simultaneously using available reference samples before the traditional sequential reconstruction process completes. This preliminary prediction approach reduces the waiting time and processing latency while maintaining the required prediction precision.
2Productivity
If multiple sub-partition based prediction is implemented simultaneously, then processing efficiency and productivity improve, but device complexity increases
Solution Approach 1:
The patent segments the prediction process into multiple independent sub-partition prediction units that can operate in parallel. Each sub-partition has its own prediction logic based on neighboring reference samples, allowing simultaneous processing and improving productivity without requiring a completely new complex system architecture.
Solution Approach 2:
The patent implements a universal prediction mechanism that can handle multiple sub-partitions using the same basic prediction logic and reference sample approach. This multi-functional prediction system processes different sub-partitions simultaneously using consistent rules, improving efficiency while keeping the core system design relatively simple and manageable.
Data Source
AI summary
Intra sub-partitions (ISP) may be enabled for a current block, for example, based on an ISP indication. The block may be partitioned into multiple sub-partitions, and a sub-partition may belong to a prediction unit (PU). A sub-partition width for the current block and a minimum prediction block width may be obtained. A PU corresponding to a current sub-partition may be determined based on the sub-partition width and the minimum prediction block width. For example, when the sub-partition width is less than the minimum prediction block width, the PU may include multiple sub-partitions. In examples, the minimum prediction block width may be four samples. Reference samples may be determined, and the PU may be predicted using the reference samples.